Leviathan Systems Public Content API

Choose your integration

Use the REST API directly, connect an MCP client, or install a JavaScript/Python client or CLI from this website. These interfaces retrieve published information about Leviathan Systems; they do not provision infrastructure or change project records.

Browse or search the resource catalog

The catalog lists published article titles and the current physical service listings. Use kind=article or kind=service, a title query in q (up to 200 characters), and a limit from 1 to 50. The default page contains up to 20 records.

Responses include items, total, limit, has_more, next_cursor and a complete next URL. Follow next until it is null. A cursor belongs to its query and content revision; begin a new query if it expires after a publication update.

GET /api/v1/resources?kind=article&q=nvlink&limit=10

{
  "apiVersion": "1",
  "items": [{ "id": "article:…", "kind": "article", "title": "…", "canonical": "https://www.leviathansystems.co/articles/…" }],
  "total": 1, "limit": 10, "has_more": false,
  "next_cursor": null, "next": null
}

The example illustrates the response shape. Live titles, counts and source dates come from the published content. A record includes markdown_url when a native Markdown representation is available.

Retrieve a document as JSON or Markdown

The existing document endpoints return Markdown by default. Request Accept: application/json for title, description, canonical, lastUpdated and the original markdown body. Article publication and revision dates stay with their source.

curl -H 'Accept: application/json' \
  https://www.leviathansystems.co/api/machine/home

curl -H 'Accept: text/markdown' \
  https://www.leviathansystems.co/api/content/guides/rack-and-stack/markdown

Discover full article representations from the catalog's markdown_url or the article index.

Official SDKs and command-line tool

Use the JavaScript and Python SDK documentation for direct package downloads, source code, cursor iteration and typed errors. The Leviathan Systems CLI lets scripts search and retrieve public resources without writing an integration. Read a catalog ID directly through GET /api/v1/resources/{id}.

Search through NLWeb

The public /ask endpoint implements the NLWeb 0.55 answer/failure format over the published catalog. It ranks matching title words and returns up to 10 canonical sources; it does not generate an answer or make a technical judgment. Send query.text in a JSON POST body, or use ?query=rack with GET. Request prefer.streaming: true or accept text/event-stream for finite start/result/complete events. No match produces a structured NO_RESULTS failure.

curl https://www.leviathansystems.co/ask   -H 'Content-Type: application/json'   -d '{"query":{"text":"rack integration"},"prefer":{"streaming":true}}'

Read Markdown without an extra discovery request

Send Accept: text/markdown or add ?mode=agent on the homepage, agent profile and main services, articles, guides and glossary indexes. ChatGPT-User, Claude-User and Perplexity-User retrieval clients receive these existing public Markdown representations by default when accepting HTML or a wildcard. An explicit Markdown quality preference takes precedence. Use ?mode=html to request the HTML page. Browser navigation and React requests keep their normal representation. Client identification selects a format and grants no additional access.

Download the structured catalog

The Schema.org JSON Lines feed provides published article titles, original revision dates and service names in one download. Discover it through the Schema Map advertised in robots.txt. Use its ETag with If-None-Match to avoid downloading an unchanged feed. The feed follows the draft Schema Feeds format and adds no claims beyond the source pages.

Changes published October 21, 2026

Added a read-only MCP connection and dedicated connection guide, structured catalog feed, heading-first public access guidance and this deprecation policy. Also added official downloadable JavaScript/Python clients, a CLI, signed discovery metadata, bounded NLWeb title retrieval, batch document reads and Markdown defaults for named AI retrieval clients. Version 1 JSON and Markdown document formats remain available.

Request limits and project inquiries

Content handlers enforce a best-effort burst budget of 300 requests per 60 seconds per server instance. This is a shared instance budget, not an account quota or a deployment-wide guarantee. No visitor identifiers are stored for this counter. RateLimit, RateLimit-Policy and compatibility limit/remaining/reset headers describe the current instance budget; reset values are seconds until reset.

Retrieve documents sequentially and honor 429 responses. The content API describes published information. Use Contact to discuss project scope, availability and commercial terms with the team.